Endgame · agentic context layer
1 / 3

01 Round two · the expanded pool, head-to-head

Twenty-four harder questions, both arms, one model.

The same twenty-four sealed questions, run on Data 360 alone and on Data 360 + Endgame — Sonnet 5 medium on both arms, one verbatim message per question, first answer graded against the same sealed key. Reported by question type, never summed. Every sealed wording, with both arms’ grading notes, is verbatim in the ledger below.

Conversational · 12 run
D360 alone7 / 12
+ Endgame9 / 12
D360: 2 partial, 3 wrong · Endgame: 1 partial, 2 wrong — the Wiz champion miss is a filed bug (END-8310); CrowdStrike is a shared miss.
Structured · 4
D360 alone4 / 4
+ Endgame4 / 4
A tie — every numeric exact on both arms.
Negative control · 1
D360 alone1 / 1
+ Endgame1 / 1
Both arms declined to fabricate.
Aggregate sweeps · 3
D360 alone3 / 3
+ Endgame0 / 3
Data 360 sweeps it. All three Endgame misses are filed: END-8311, END-8313, END-8314 — rerun candidates, not guarantees.
Conversational12 · D360 7 pass · Endgame 9 pass
questiond360 alone+ endgame
CrowdStrike renewal date
Illumio worksheet due
Hugging Face renewal
W&B last contact
Stripe budget owner
Wiz champion bug filed · END-8310
Schneider hold-up
Klarna’s last ask
Stripe owed deliverable
Databricks real blocker
Capital One conditions
Nike pressure-tests
Aggregate sweeps3 · D360 3 pass · Endgame 0 pass
questiond360 alone+ endgame
Quiet $1M+ accounts bug filed · END-8311
Wrong-domain contacts gap filed · END-8313
Competitive loss gap filed · END-8314
Negative control1 · both arms pass
questiond360 alone+ endgame
Slipped-deals probe
Structured4 · every numeric exact, both arms
questiond360 alone+ endgame
Pipeline by stage
Top five deals
2025 bookings
Q4-2026 pipeline
Held-out + variant4 · reported separately
questiond360 alone+ endgame
Wells Fargo last talk
Snowflake hold-up expected miss · documented pre-run
MongoDB loss lessons
Quiet $500k+ accounts variant

passpartialfail

both arms recorded 2026-08-01 · sonnet 5 medium · claude desktop · same sealed key · round-one carried cells not re-run: endgame 3/3, data 360 0/3 · internal working copy — carries linear ids, not for the packet

© 2026 Endgame Labs, Inc. · Confidential & Proprietary

2 / 3

02 The comparison ledger

Every run, both arms, on the record.

The sealed wording of each question, the verdict on each arm, and what each answer did. Round one’s three carried cells are not re-run: Endgame 3/3, Data 360 0/3.

Conversational · 12 run this pass · D360 7 pass, 2 partial, 3 wrong · Endgame 9 pass, 1 partial, 2 wrong

d360
alone
+ end
game

“When does the CrowdStrike renewal actually come up?”

+EGReported the CRM’s December 3 close date; the customer’s corrected November 3 date never surfaced.

D360Gave the CRM’s December 3 date — trusted the close-date field, never queried email; the corrected November 3 date never surfaced.

“On Illumio — they were going to fill in the sizing worksheet. When was that actually due, and did it ever come back?”

+EGDue April 5, and it never came back — the call, the ask, and the overdue status all cited.

D360Echoed “Waiting on legal” and claimed zero Illumio engagement; the call holding the answer (due April 5, never returned) was never read.

“What’s the Hugging Face renewal worth, and when is it actually closing?”

+EG$510,000, closing September 7 — both exact.

D360$510,000, closing September 7 — both exact.

“Weights & Biases is supposed to close this month. When did we last actually hear from them?”

+EGThe June 12 email, its sender, and all three unanswered asks — exact.

D360June 12 inbound from Kiran Okonkwo, nothing since — sealed date and sender exact.

“On the Stripe tracing deal — who owns the budget, and are they attached to that opportunity?”

+EGNamed the budget owner in her own words — and that she is not attached to the tracing deal. The wrong-deal trap, avoided.

D360Named the budget owner but ruled from the role table — “attached” via the Services deal, the sealed trap; never bound tracing to the deal where she has no role.

“Who’s our champion at Wiz, and how do I reach them?”

+EGNamed the wrong person as champion; the parent-company email address that answers the question never appeared. Root cause, verified live: the graph holds the champion role on both Wiz deals; the person-view tool payload strips role edges — filed as END-8310.

D360Rania Hansen, VP Engineering, with the parent-domain address — the sealed champion, exact.

“Schneider Electric — what’s actually holding this up?”

+EGThe missing volume figure, both unanswered asks, and the customer’s mirrored request — exact.

D360Inverted the hold-up — said we owe an ROI summary; the sealed blocker is their ballpark volume figure, asked twice and never answered.

“What’s the last thing Klarna asked us for?”

+EGJune 21, the asker, and the exact deliverable — flagged as the blocker, with no response since.

D360The pricing model with tracing volumes broken out, quoted verbatim with its asker and date; July 2 correctly read as no new ask.

“What is Stripe still waiting on from us, and how long have they been asking?”

+EGThe deliverable and the July 2 repeat ask exact; the first ask misdated, and one timeline entry unverifiable against the sealed record.

D360Right deliverable, but dated from calls (“since Jan 19”); both sealed email asks (June 20, July 2) missed.

“What’s actually blocking the Databricks expansion?”

+EGThe June 12 Slack pushback, by name — a fact that exists nowhere in the CRM.

D360Quoted the June 12 Slack pushback — per-host pricing vs Datadog spend, named as the real blocker — the Slack-only objection recovered.

“Capital One is a $1.5M deal still in Discovery. What has their CTO told us she needs before she’ll sign?”

+EGNamed her as the budget owner with one sealed condition verbatim; one of four sealed conditions recalled.

D360Named the CTO; sealed conditions recalled 5 of 5.

“Nike’s a $1.3M deal at Proposal. What does Tariq think stops it, and what’s he pressure-testing?”

+EGBoth sealed blockers, with the procurement and pricing quotes verbatim.

D360Procurement-gauntlet and per-unit-vs-Datadog quotes verbatim — both sealed halves exact.

passpartialfail

© 2026 Endgame Labs, Inc. · Confidential & Proprietary

3 / 3

03 The comparison ledger, continued

The ledger, continued.

Aggregate sweeps · 3 questions · Endgame 0 correct

d360
alone
+ end
game

“Which accounts with over a million dollars of open pipeline haven’t had a real conversation in more than a year?”

+EGAnswered “none” — trusted the activity rollup the question targets; none of the four members found. Root cause, verified live: the served recency field is a CRM activity rollup with no provenance flag — filed as END-8311; the correct dates are present in the graph (the $500k-variant run found all four).

D360All four sealed members with exact last-conversation dates; never-engaged tier flagged separately.

“Do we have any contacts whose email is on a different company’s domain than the account they belong to?”

+EGSampled ~45 contacts, said so, and offered a full sweep — but found neither sealed member. Attribution: a capability gap, filed as END-8313 — no org-wide domain-comparison primitive on the tool surface; a derived domain-match flag is proposed for a one-call answer. The underlying graph data is correct, both members verified present.

D360Full 1,395-contact scan; exactly the two sealed mismatches found.

“Have we lost anything to a competitor, and what was the stated reason?”

+EGAnswered “no stated reason recorded” — the reason sits in the lost deal’s own description field. Attribution: two capability gaps, filed together as END-8314 (flagged separable) — fact search rejects opportunity-scoped filters, and description text is not extracted into the fact layer; the account-scoped route reaches the same fact (see the MongoDB row).

D360The exact deal, amount, date, and verbatim recorded reason; the competitor’s name not stated.

Negative control · 1 question

“Which of my biggest deals keep moving their close date? I want the ones that have slipped three times or more.”

+EGStated the history isn’t tracked, declined to build the list, and distrusted the templated slip language it found.

D360Checked push counts and field history, stated the movement isn’t tracked, named no deals — no fabrication.

Structured · 4 questions · Endgame every numeric exact

“How much open pipeline do we have, and how does it break down by stage?”

+EG$169,055,000 across 259 open deals, with the full stage split — six of six figures exact.

D360$169.1M across 259 and all five stage rows exact — six of six.

“What are our five largest open deals?”

+EGAll five names and amounts exact, in the correct order.

D360All five names and amounts exact, correct order.

“What did we close-win in calendar 2025 — total value and deal count?”

+EG$58,865,000 across 86 — both exact.

D360$58,865,000 across 86 — both exact.

“How much pipeline is scheduled to close between October 1 and December 31, 2026?”

+EG$101,520,000 across 158 — both exact.

D360$101.52M across 158 — both exact.

Held-out · 3 questions · reported separately

“The Wells Fargo renewal is a $1.6M deal in negotiation. When did we last talk to them, and how did that conversation end?”

+EGOctober 12, 2024, the sender, the unanswered cost-model ask — and the nearly two-year silence flagged.

D360October 12, 2024, Jamie Mensah, ended on the unanswered cost-model ask — date and ending exact.

“The Snowflake logs expansion has been at Proposal for a while. What’s the actual hold-up?”

+EGExpected miss — returned the story the serving layer carries; the gap was documented before the run, and the question held out for exactly that reason.

D360Declined to query — zero tool calls; read “at Proposal” as a project tracker and asked where it’s tracked.

“We lost a platform deal at MongoDB. What happened, and are we repeating it on the current motions?”

+EGThe exact deal, date, and amount — and the repeating pattern mapped onto the live deals.

D360Loss, date, amount, verbatim reason, and the repeat pattern answered with specifics.

Variant · 1 question

“Which accounts with over $500,000 of open pipeline haven’t had a real conversation in more than a year?”

+EGFound part of the member set, plus a correct never-engaged tier — and said exactly how far its sweep went.

D360All four sealed members with exact dates inside a 78-account sweep; aged-out vs never-engaged distinguished.

Identity questions · 2 · identity unresolved on both arms · outside the 24

“Which of my million-dollar-plus open deals have only one person attached?”

+EGExcluded — the demo connection carries no acting user, so “my” cannot resolve. Distinct from failed.

D360Identity unresolved — silently scoped to the whole org; the ten deals returned match the sealed org-wide set exactly. Outside the score.

“Which of my open deals over $500,000 have only one person attached?”

+EGExcluded — same identity gap.

D360Identity unresolved — silently scoped org-wide; the count of 48 matches the sealed set exactly. Outside the score.

passpartialfailexcluded, not scored

grades against the sealed key · one verbatim message per question, first answer · sonnet 5 medium, both arms · endgame arm 2026-08-01 · data 360 arm pending

© 2026 Endgame Labs, Inc. · Confidential & Proprietary

Endgame · round two, both arms · © 2026 Endgame Labs, Inc. · Confidential & Proprietary